Article Text

Quadriceps or hip exercises for patellofemoral pain? A randomised controlled equivalence trial
  1. Rudi Hansen1,2,
  2. Christoffer Brushøj2,
  3. Michael Skovdal Rathleff3,4,
  4. S Peter Magnusson2,5,
  5. Marius Henriksen6
  1. 1 Department of Physical and Occupational Therapy, Copenhagen University Hospital - Bispebjerg and Frederiksberg, Copenhagen, Denmark
  2. 2 Institute of Sports Medicine Copenhagen, Department of Orthopedic Surgery, Copenhagen University Hospital - Bispebjerg and Frederiksberg, Copenhagen, Denmark
  3. 3 Center for General Practice, Aalborg University, Aalborg, Denmark
  4. 4 Department of Health Science and Technology, Aalborg University, Aalborg, Denmark
  5. 5 Center for Healthy Aging, Department of Clinical Medicine, University of Copenhagen, Copenhagen, Denmark
  6. 6 The Parker Institute, Copenhagen University Hospital - Bispebjerg and Frederiksberg, Frederiksberg, Denmark
  1. Correspondence to Mr Rudi Hansen, Department of Physical and Occupational Therapy, Bispebjerg Hospital, Copenhagen, Denmark; rudi.hansen{at}


Objective To assess effectiveness equivalence between two commonly prescribed 12-week exercise programmes targeting either the quadriceps or the hip muscles in patients with patellofemoral pain (PFP).

Methods This randomised controlled equivalence trial included patients with a clinical diagnosis of PFP. Participants were randomly assigned to either a 12-week quadriceps-focused exercise (QE) or a hip-focused exercise (HE) programme. The primary outcome was the change in Anterior Knee Pain Scale (AKPS) (0–100) from baseline to 12-week follow-up. Prespecified equivalence margins of ±8 points on the AKPS were chosen to demonstrate comparable effectiveness. Key secondary outcomes were the Knee Injury and Osteoarthritis Outcome Score (KOOS) questionnaire pain, physical function and knee-related quality of life subscales.

Results 200 participants underwent randomisation; 100 assigned to QE and 100 to HE (mean age 27.2 years (SD 6.4); 69% women). The least squares mean changes in AKPS (primary outcome) were 7.6 for QE and 7.0 for HE (difference 0.6 points, 95% CI −2.0 to 3.2; test for equivalence p<0.0001), although neither programme surpassed the minimal clinically important change threshold. None of the group differences in key secondary outcomes exceeded predefined equivalence margins.

Conclusion The 12-week QE and HE protocols provided equivalent improvements in symptoms and function for patients with PFP.

Trial registration number NCT03069547.

  • Knee injuries
  • Randomized Controlled Trial
  • Exercise Therapy
  • Rehabilitation

Data availability statement

Data are available upon reasonable request.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • Current evidence supports exercise therapy in the treatment of patients with patellofemoral pain. However, there is uncertainty about the comparative effectiveness of hip and knee exercises and high-quality evidence is needed to guide clinical practice.


  • This study demonstrates that quadriceps-focused exercises and hip-focused exercises provide equivalent benefits for patients with patellofemoral pain, but the improvement did not reach the established minimally clinical important change threshold.


  • Based on this study, clinicians can include patient preferences and individualisation in the choice of either hip or knee focused exercises in the management of patients with patellofemoral pain.


Patellofemoral pain (PFP) is a common knee problem, with point prevalence from 6% to 7% in adolescents and up to 13% in young adults.1–3 More than one in two with PFP report persistent pain after 5–8 years,4 with an associated frequent use of painkillers, a lower physical activity level and low quality of life.1 4–6 Recent systematic reviews and a network meta-analysis recommend exercise therapy (mainly comprising exercises for the hip, the knee or both the hip and knee) for improving pain and function in people with PFP.7–10 However, these studies also underline the uncertainty about which type of exercises that are most effective for PFP. Despite the latest consensus document on managing PFP recommends including hip exercises, direct comparisons of exercise protocols are few,9 with short intervention and follow-up periods,11 and with sample sizes insufficient to detect clinically relevant differences in outcomes.12 13 Collectively, this challenges the choice of the most appropriate treatment and may also explain the variation in clinical practice.14 Hence, there is a need for large high-quality studies of comparative effectiveness of quadriceps and hip muscle exercises for PFP. Accordingly, the aim of this study was to assess effectiveness equivalence between a focused ‘Quadriceps Exercise’ (QE) protocol and a focused ‘Hip Exercise’ (HE) protocol on symptoms and function in patients with PFP.


Study design

In this single-centre randomised, controlled, assessor-blinded, equivalence trial with two parallel intervention groups we compared a QE and HE protocol. Evaluations and assessments took place at the Department of Physical and Occupational Therapy at Bispebjerg-Frederiksberg Hospital, Copenhagen, Denmark, at baseline and 12 weeks. Further, the participants were invited to an online collection of patient-reported outcomes 26 weeks after baseline. The trial design is illustrated in the online supplemental file. This report follows the CONSORT (Consolidated Standards of Reporting Trials) extension for non-pharmacological treatments guideline and the Template for Intervention Description and Replication (TIDieR) checklist for intervention description.15 16 The study was registered prospectively at on 3 March 2017.

Supplemental material

Patient and public involvement statement

Patients were not engaged in the development stages of the study nor in the conduct or oversight of the study. All participants were offered a lay language resume of results and conclusion of the study by email.


Between 10 April 2017 and 3 December 2021, participants were recruited from the Institute of Sports Medicine Copenhagen (ISMC), Bispebjerg-Frederiksberg Hospital, Copenhagen, Denmark. Inclusion was halted for 8 weeks from 12 March 2020 due to the COVID-19 pandemic. ISMC is a medical unit for patients with injuries in the musculoskeletal system caused by participation in sports activities. Patients are referred to ISMC from primary care physicians, and from ISMC to specialised rehabilitation at the Department of Physical and Occupational Therapy. All participants underwent a clinical examination by a specialist in sports medicine but were not screened for eligibility using radiographs or other imaging. All participants provided written informed consent before participation.

Inclusion criteria were a clinical diagnosis of PFP in at least one knee confirmed by an experienced sports medicine physician, average knee pain during activities of daily living in the last week of ≥3/10 on a verbal rating scale, insidious onset of symptoms unrelated to trauma, persistent pain for at least 4 weeks and anterior knee pain associated with at least three of the following: During or after activity, prolonged sitting, stair ascent or descent or squatting. The exclusion criteria were other knee conditions, including meniscal or other intra-articular injuries to the knee, history of recurrent patellar subluxation or dislocation and previous knee surgery. Potential participants were informed about the trial during an interview with a sports medicine physician, and after at least 24 hours of consideration an investigator obtained written informed consent and coordinated trial visits. The most symptomatic knee at baseline was chosen as the study knee.

Randomisation and blinding

Before randomisation, demographic information and all baseline measures were obtained. Participants were randomly assigned (1:1) in permuted blocks of 4 and 6 (randomly distributed) according to a computer-generated list of random numbers, to one of the two groups (QE or HE). Individual allocations were concealed in sealed opaque envelopes, stored in a locked cupboard without access for investigators or outcome assessor and delivered sequentially to the study physiotherapist at randomisation. The physiotherapists delivering the interventions and participants were not blinded to treatment allocation. The investigators and the outcome assessor were blinded to allocation, and participants were requested not to disclose allocation during clinical assessments.


Both the hip and knee focused exercise programmes were inspired by previous research17 and followed recommended prescribing guidelines.18 19 The exercise interventions lasted for 12 weeks with 3 weekly home-based exercise sessions consisting of three sets of 8–12 repetitions. Key parameters of the exercise programmes are shown in table 1, and the complete description of the interventions is provided in the online supplemental file.

Table 1

Key parameters of the exercise programmes

The hip exercise (HE) programme consisted of hip external rotation (clam shell), side-lying/standing hip abduction and prone/standing hip extension. The HE exercises were chosen due to their documented activation of the hip abductors, external rotators and hip extensors,20–23 wide use in clinical practice, and because they do not strain the patellofemoral joint excessively. The quadriceps exercise (QE) programme consisted of sitting knee extension, squat and forward lunge. The exercises have been shown effective in recruiting the quadriceps muscle22 and appear effective in the treatment of PFP.24 25

Both exercise programmes were initiated at an individual clinical visit. An experienced physiotherapist introduced the participant to the allocated exercise programme (QE or HE) and provided instructions to the individual exercises. Elastic bands, free weights and body weight were used to provide resistance. The participants were informed to perform 8–12 repetitions in each set. The last repetitions should be difficult to perform while still allowing the participant to maintain high quality of movement (ie, full range of motion and without any compensatory movements (judged by the physiotherapist)) throughout the entire programme. The participants were instructed to increase resistance whenever they could complete 14 repetitions in a set. This was emphasised during the instructional session and during each follow-up visit19 (progression principles are specified in the online supplemental file). To reflect normal clinical practice, the exercise programmes included monthly clinical supervision visits. Reduction in the exercise load (Range of Motion (ROM), number of sets/repetitions) could be made in case of significant knee pain exacerbations.

All participants—irrespective of group allocation—received the information leaflet ‘Managing my patellofemoral pain’ containing general information on possible causes and management of PFP. The leaflet is available in the online supplemental file. Further a comprehensive exercise leaflet with guidance on the exercises, progression/regression and pain management was handed out. All the physiotherapists involved in the study (n=5) were instructed to communicate in the same way, and training sessions were held in the planning stage to ensure standardisation of communication and practice.

Adherence to the prescribed exercise protocol was monitored by a self-administered exercise diary, which the participants were encouraged to fill in after each exercise session. The participants were asked to record the date, number of repetitions and sets for each exercise and the resistance (ie, elastic band colour corresponding to a specified resistance or weights in kg). The exercise diary was brought at the monthly clinical visits to optimise compliance and handed in at the 12 weeks assessment. The criteria for satisfactory intervention adherence in both groups was 24 of the 36 scheduled training sessions (66%).

Primary outcome

The primary outcome was change from baseline in the Anterior Knee Pain Scale (AKPS) questionnaire at week 12. The AKPS questionnaire is a widely used and well-validated questionnaire for assessing the severity of symptoms and physical limitations in people with PFP.26 The 13 items in the questionnaire are summed up to give a total score ranging from 0 to 100, with high scores indicating less symptoms. The minimal clinically important change is established at 8–10 points.27

Key secondary and other secondary outcomes

Key secondary outcomes were changes from baseline in the Knee Injury and Osteoarthritis Outcome Score (KOOS) questionnaire pain, physical function and knee-related quality of life subscales.28 Other secondary outcomes included changes from baseline in the KOOS sports/recreation and symptoms subscales, Pain Self-Efficacy Questionnaire,29 the EuroQoL EQ-5D-3L Questionnaire,30 assessment of pain on a 0–10 Numeric Rating Scale (NRS) during activity (30 s of performing repeated deep knee-bends from a standing position)31 and global perceived effect on overall health, pain and function measured on a 15-point Likert scale ranging from −7 (much worse) to +7 (much better). Further, isometric muscle strength of hip abductors, hip adductors, hip external rotators, hip internal rotators, hip extensors, hip flexors, knee flexors and knee extensors were measured with a handheld dynamometer. The testing was performed in a clinical examination room with the participant lying or sitting on an examination table with and without external fixation according to published and validated protocols.32–34 Three consecutive isometric maximal contractions were performed with a 30 s rest period between each trial, and the maximum value was used for analysis. Changes from baseline in the patient reported outcomes (questionnaires) at week 26 were also recorded. The physiotherapists performed an estimation of each participant’s prognosis just after the initial instructional session for the purpose of a secondary analysis on outcome prediction (to be reported in a separate paper; study protocol available on request).

Sample size

The sample size was calculated to allow for test of equivalence of the treatment groups at 90% power and an alpha level of 0.05 using a two one-sided test (one-sided alpha of 0.025) with equivalence margins of ±8 AKPS points, assuming a mean difference of 0 points and a common SD of 15 points.35 36 From this, 77 patients were required in each treatment group. To account for a dropout rate of approximately 20% the sample size was a priori increased to 100 participants in each group.

Statistical analysis

The primary analysis was performed using the intention-to-treat (ITT) population; patients were assessed and analysed as members of their randomised groups, irrespective of adherence to the planned course of treatment. Continuous outcomes were analysed as change from baseline using repeated measures linear mixed models with group (two levels), time (two levels; week 12 and 26) and the corresponding interaction as fixed effects and participants as random effects (normal distribution assumed). Adjustments were made for baseline values. Assumptions underlying the linear mixed models were assessed by visual inspection of Q-Q plots and residual plots (for normality of residuals and homogeneity of variance, respectively), and plots of quantitative predictors against residuals (for assessment of linearity of covariates). The assumptions were judged as fulfilled. Results are reported as least squares means and differences between least squares means with two-sided 95% CIs. The group difference in the primary outcome was assessed for equivalence by a two one-sided test of equivalence with alpha 0.025 assessing if the 95% CI respects the predefined equivalence margin of ±8 AKPS points corresponding to the established cut-off value for making the distinction between improved or unimproved.27 No explicit adjustments for multiplicity were applied, rather the key secondary outcome measures were analysed in a prioritised order. Missing values for items in the AKPS and Pain Self-Efficacy Questionnaire were substituted with the arithmetic mean of values from the available items. If more than 25% of items were missing, the outcome was regarded as missing for the patient.13 37 For the KOOS questionnaire, a mean score for each subscale was calculated, as long as at least 50% of the items were answered for each subscale. If more than 50% of the subscale items were omitted, the response was considered invalid. Imputation of missing item values for the EuroQoL EQ-5D-3L Questionnaire was handled according to the user guide ( Imputation of missing values in AKPS constituted less than 5% of all questionnaire data. Complete missing data were handled implicitly in the intention-to-treat analysis by the linear mixed models.38 Sensitivity analyses were performed for the primary and key secondary outcomes by repeating the primary analyses on the per-protocol population predefined as participants with satisfactory adherence and without major protocol deviations.39 Further, we performed generalised estimating equation analyses of the primary and key secondary outcomes on the ITT population with missing data handled by inverse probability weighting with weights estimated from a logistic regression model for predicting missingness, adjusted for the baseline value. Finally, we performed an analysis of covariance of the primary and key secondary outcomes at week 12 (ie, without the repeated measures) on the ITT population with missing data replaced using multiple imputation (100 multiply imputed data sets with missing data predicted using baseline data) adjusted for the baseline value. For the sensitivity analysis the underlying assumptions were assessed in the same way as the primary analysis and the assumptions were judged as fulfilled. If the primary analysis and the sensitivity analyses confirm each other, confidence in the results is increased both regarding equivalence and superiority claims. All analyses were performed in SAS V.9.4 (SAS Institute, Cary, North Carolina, USA).

Equity, diversity and inclusion statement

The study included individuals with knee pain referred from primary care to specialised rehabilitation. The study population included participants from a broad range of ethnic/racial and socioeconomic backgrounds. The research team included five men and four women. The author team included one junior scholar (man), one senior clinician (man) and three senior academics (all men).



From 10 April 2017 through 3 December 2021, 288 individuals were screened for eligibility (figure 1); 88 were ineligible for inclusion. Thus, 200 participants underwent randomisation; 100 were assigned to QE and 100 to HE. The mean age was 27.2 years (SD 6.4); 69% were women; and the mean BMI was 22.6 (SD 3.0). Baseline characteristics were similar in the two groups (table 2). Participants completed on average 28 (77%) training sessions out of 36 possible sessions. During the course of the intervention, six participants (four QE and two HE) had alterations to their allocated exercise programmes (mainly reduced ROM in the weight bearing exercises) and eight (five QE and three HE) had the number of sets and repetitions reduced due to knee pain exacerbations.

Figure 1

Consolidated Standards of Reporting Trials flow diagram.

Table 2

Demographics and baseline characteristics

Primary outcome

The mean changes in AKPS questionnaire score from baseline to week 12 were 7.6 (95% CI 5.8 to 9.5) in the QE group and 7.0 (95% CI 5.2 to 8.9) in the HE group (group difference: 0.6 points, 95% CI −2.0 to 3.2; p=0.636 for test of superiority). The 95% CI of the group difference in change in AKPS questionnaire from baseline to week 12 was within the predefined equivalence margin of ±8 points (p<0.0001 for equivalence, table 3). The trajectories of the AKPS questionnaire are shown in figure 2.

Table 3

Primary and secondary outcomes at week 12 in the intention-to-treat population. Based on repeated measures linear mixed models, where missing data is assumed to be missing at random

Figure 2

Trajectories of the Anterior Knee Pain Scale (AKPS) questionnaire in the intention-to-treat population. High values represent high levels of self-reported function; low values represent low levels of self-reported function. Data points represent least squares means; error bars represent 95% CI.

Key secondary and other secondary outcomes

For the key secondary outcomes, the estimated treatment differences between groups at week 12 were 2.9 points (95% CI −0.6 to 6.5) for KOOS pain score, 1.0 points (95% CI −1.8 to 3.9) for KOOS function and −1.2 points (95% CI-6.1 to 3.7) for KOOS quality of life score. The key secondary outcomes all respected the predefined criteria for equivalence (table 3). Finally, the results in the primary and key secondary outcomes appeared unchanged at week 26 (table 4). There were no statistically or clinically significant differences between groups in the other secondary, safety and exploratory outcomes at week 12 (table 3) and week 26 (table 4). The overall pattern of results for all outcomes was unchanged in the sensitivity analyses (online supplemental tables S1-S3).

Table 4

Primary and secondary outcomes at week 26 in the intention-to-treat population. Based on repeated measures linear mixed models, where missing data is assumed to be missing at random


Adverse events were typically mild to moderate, mostly related to muscle soreness, and were similar in the two groups (table 5). Severe adverse events that gave interference with the participants’ usual activities were exacerbation of knee pain (n=2), headache (n=1) and back pain (n=1).

Table 5

Adverse events in the intention-to-treat population


The results of this study provide much needed evidence to inform clinical practice and highlight that an exercise programme that focused on either quadriceps or hip muscles provided equivalent improvements in symptoms and function in the short (12 weeks) and medium term (26 weeks). Treatment adherence was similar in the two groups as were adverse events that were few.

Our results support recently published randomised controlled trials comparing hip and knee focused exercise protocols. In Hott et al,13 112 patients were randomised to three groups (a 6-week intervention consisting of patient education combined with isolated hip-focused exercise, traditional knee-focused exercise or free physical activity); the data indicated no difference in the primary outcome AKPS between groups. This is in line with previous studies, showing no difference in pain and function at 6–8 weeks between a hip and a knee exercise group.40 41 Three studies have found hip exercises to be more effective than knee-focused exercise42–44; however, the sample sizes were typically quite modest (15–18 per group), and one study lacked randomisation. This study is the first using an equivalence design that allows us to draw reliable conclusions regarding the comparative effectiveness of hip and knee focused exercises for PFP. Hence, our results extend current understanding and effectively demonstrate equivalent effectiveness of hip and knee focused exercise for PFP.

Both exercise programmes were associated with improvements in AKPS score (7.6 points for QE and 7.0 points for HE), but the improvements did not surpass the minimal clinically important change. The within-group changes for QE and HE are similar to those previously reported,13 but are somewhat lower than those reported in other randomised controlled trials evaluating the effect of hip and knee exercises in adolescents and adults with PFP.35 40 41 45–47 This difference may be explained by the setting of this study. Patients included in this study were referred to specialised rehabilitation most often due to long-standing symptoms, which is reflected in the patient demographics. Previous studies have shown that long symptom duration is associated with worse outcomes (irrespective of treatment)4 48 49 which may explain the somewhat small within group changes. Mean pain duration in this study was higher when compared with most studies that report on this.40 45 47 Another plausible explanation for the small within group changes could be differences in attention and supervision during the intervention period compared with other studies. Most of the interventions in comparable studies were supervised, but this is not always feasible in a clinical setting. The patient–physiotherapist relationship and the overall healthcare setting are relevant categories of contextual factors that may modify treatment effects.50 Lastly, the baseline AKPS scores were relatively high considering the long pain duration, which could potentially introduce a ceiling effect on individual items, which in turn could affect the change scores.

Both groups had 10–11% improvements in hip abduction and knee extension muscle strength after the 12-week training period irrespective of group allocation, which is similar to previous studies.13 40 44 Since some aspects of the hip exercises involve weight bearing, several other muscles are recruited when performing the exercise, including the quadriceps. Likewise for the quadriceps focused exercises, a possible parallel training of the hip (and other synergistic) muscles is likely to have occurred. One could argue that this may explain the lack of group difference in the outcomes. However, in a large randomised clinical trial with 218 participants with PFP, increases in muscle strength did not mediate improvements in pain.51 This suggests that improvements in muscle strength might not be the driver of beneficial outcomes, and that other mechanisms are more important.

The somewhat modest improvements seen in our and other recent studies on exercises for PFP raise the question if treatment plans focusing on strengthening and biomechanically informed movement quality alone address the right components contributing to the pain experience. Growing evidence suggests that psychological features may play a role in long-standing PFP.52–54 Future studies should aim at identifying possible patient characteristics that predict successful outcomes.

Clinical implications

We found that quadriceps exercises and hip exercises are equally effective treatments in the management of patients with PFP. However, the improvements did not reach the established minimally clinical important change threshold, and therefore training the quadriceps or hip muscles separately may not be effective in improving symptoms and function. This is supported by the most recent consensus document that recommends combining quadriceps and hip exercise.9 This may also imply that therapists should use their clinical reasoning and include patient preferences when designing an exercise rehabilitation programme for the individual patient. Such shared decision may improve healthcare efficiency and is recommended in the rehabilitation of patients with PFP.9 55 56 However, although personalisation of exercise interventions to individual patients or subgroup of patients may be a useful strategy that can ultimately lead to improved outcomes for patients,57 such strategy remains to be supported by research evidence—preferably from prospective randomised trials.

Limitations and strengths

There are inherent limitations to this study. First, the exercise programmes were home-based with limited supervision, which may introduce a risk that the exercises were not performed correctly. While more regular visits to the clinician would assure adherence and fidelity to the treatments, this would not be in accordance with our intention to resemble a clinical setting, where multiple weekly visits are not feasible.14 On the other hand, the resemblance of normal clinical practice increases external validity of our results. Second, the exercise adherence data was based on self-reporting, which introduces an inherent risk of overestimation due to social desirability, recall period and selective recall.58 Third, this study was a single centre trial which may limit the external validity. Fourth, as the main part of the interventions were unsupervised, contamination (deliberate switch of exercise programme) may have occurred. Fifth, as part of a prognostic substudy (to be published separately) the physiotherapists recorded the participants’ projected prognosis after the first clinical encounter but not disclosed to the participants. As this was done post randomisation it may have introduced some expectation biases with the physiotherapists. However, such prognoses are inherent in clinical interactions between patients and healthcare providers and thus this does not represent deviations from normal clinical practice. Further, the ITT analysis estimates the effect of assigned but not received treatment and may be biased by non-adherence. Also, missing outcome data can introduce selection bias even when missingness in the two treatment groups are similar. Finally, the primary analysis assumes data missing at random, which may not be true. However, the sensitivity analyses confirm the primary analyses increasing the confidence in the conclusions. The strengths of this trial included the relatively large sample size and the equivalence design, which increased the precision of the estimated group differences, and the reporting of adverse events. Furthermore, this is the first study comparing hip and knee exercises with an intervention period of 12 weeks, with comparable studies ranging from 3 to 8 weeks of intervention.10 11


In individuals with PFP, 12-week quadriceps-focused and hip-focused exercise programmes provided equivalent effectiveness for improvements in symptoms and function. More research is required to define personalised or combined exercise programmes with greater effectiveness.

Data availability statement

Data are available upon reasonable request.

Ethics statements

Patient consent for publication

Ethics approval

This study involves human participants and was approved by Health Research Ethics Committee of Capital Region Denmark, number: H-16045755 (approved 15 December 2016). Participants gave informed consent to participate in the study before taking part.


We thank the participants who volunteered to the study, and the physiotherapists at the Department of Physical and Occupational Therapy at Bispebjerg-Frederiksberg Hospital who delivered treatments.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Twitter @rudihansen76

  • Contributors RH: concept/design, data collection, data analysis, manuscript draft, critical revision. MH: concept/design, data analysis, critical revision, guarantor. NCB: concept/design, medically responsible, eligibility screening, critical revision. MSR: concept/design, critical revision. SPM: critical revision.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.